PDF to Text & Markdown Extractor for RAG & AI | $3/1K PDFs avatar

PDF to Text & Markdown Extractor for RAG & AI | $3/1K PDFs

Pricing

from $2.40 / 1,000 pdf extracteds

Go to Apify Store
PDF to Text & Markdown Extractor for RAG & AI | $3/1K PDFs

PDF to Text & Markdown Extractor for RAG & AI | $3/1K PDFs

Extract clean text and Markdown from PDF URLs for RAG, AI agents and data pipelines. Per-page text, metadata (title, author, dates), word counts and scanned-PDF detection. No proxy, no API key. $3 per 1,000 PDFs; failed and image-only PDFs are free.

Pricing

from $2.40 / 1,000 pdf extracteds

Rating

0.0

(0)

Developer

Impakt

Impakt

Maintained by Community

Actor stats

0

Bookmarked

2

Total users

1

Monthly active users

16 hours ago

Last modified

Features

Limited permissions

Share

PDF to Text & Markdown Extractor: Clean Text from PDF URLs for RAG, AI Agents & Pipelines

Try it in one click: press Start with the two prefilled PDFs (a US tax form and the "Attention Is All You Need" paper) and get clean text and Markdown in about 5 seconds. $3 per 1,000 PDFs. PDFs that fail to download or have no text layer are free.

Turn PDF links into clean, structured text. Give the Actor a list of PDF URLs and it downloads each file, reads the embedded text layer and returns plain text, Markdown (headings and bullet lists detected from font sizes), optional per-page text, metadata (title, author, creator, creation and modification dates, PDF version), page and word counts, and a SHA-256 hash of the file.

It is built for RAG pipelines, AI agents and LLM workflows that need reliable document text, and for research, compliance, finance and data teams that process reports, filings, papers, invoices, manuals and contracts at scale. Everything is plain HTTP plus a PDF parser: no browser, no proxy, no API key.

What you get

FieldDescription
url, status, errorSource link and outcome (ok, no_text_layer, not_a_pdf, download_failed, too_large, password_protected, invalid_pdf, invalid_url)
textExtracted plain text, paragraphs separated by blank lines
markdownMarkdown with #/##/### headings and - bullets
pagesOptional array of {page, text} for citations and chunking
title, author, subject, creator, producerPDF metadata when present
createdAt, modifiedAt, pdfVersionDates in ISO format
pageCount, pagesExtracted, truncatedSize of the document and how much was processed
wordCount, charCountQuick size metrics for chunking and cost estimates
needsOcrtrue when the PDF has almost no embedded text (a scan)
fileSizeBytes, sha256, processingMsUseful for caching and de-duplication

Input example

{
"pdfUrls": ["https://arxiv.org/pdf/1706.03762"],
"outputFormat": "both",
"includePages": true,
"maxPagesPerPdf": 50
}

Output example (shortened)

{
"url": "https://arxiv.org/pdf/1706.03762",
"status": "ok",
"title": null,
"pageCount": 15,
"pagesExtracted": 15,
"wordCount": 5879,
"needsOcr": false,
"markdown": "# Attention Is All You Need\n\nAbstract\n\nThe dominant sequence transduction models are based on complex recurrent or convolutional neural networks ...",
"sha256": "..."
}

Pricing

Pay per event, no subscription:

EventPrice
Actor start$0.0001
Each PDF extracted$0.003 (about $3 per 1,000 PDFs, discounts on paid Apify plans)

You are not charged for failed downloads, invalid files, password-protected files, oversized files, or PDFs without a text layer. Use Max PDFs and Apify's maximum charge per run to cap spending.

Use cases

  • RAG and AI agents: convert documents to Markdown before chunking and embedding. Call the Actor through the Apify API or MCP server.
  • Finance and compliance: extract text from filings, reports and policies for search or review.
  • Research: build corpora from open-access papers and public reports.
  • Document pipelines: pair with the Apify scheduler and webhooks to process new PDFs automatically, and keep the sha256 to skip duplicates.

How it works

  1. Each URL is validated (only public http/https addresses are allowed) and downloaded with size limits and retries.
  2. The file is parsed with Mozilla's pdf.js, and text items are grouped into lines.
  3. Line font sizes are compared with the document's body font to detect headings. Bullets and hyphenated line breaks are handled, and rotated margin text (such as arXiv stamps) is ignored.
  4. Results are returned as one dataset item per PDF.

Honest limits

  • No OCR. Scanned PDFs (images of text) are detected and returned as no_text_layer free of charge. Use an OCR tool for those.
  • Markdown is heuristic. Headings and lists come from font sizes, so unusual layouts can differ. Tables are returned as text, not Markdown tables. Multi-column pages are read in vertical order and may interleave columns.
  • Password-protected PDFs are not opened.
  • Only the first Max pages per PDF pages are processed (100 by default). Larger files than Max file size are skipped.
  • Some servers block automated downloads; those links are reported as download_failed.

Legality

The Actor only downloads the files you list. You are responsible for having the right to process those documents and for following copyright, privacy and data-protection laws that apply to them. The Actor does not crawl websites and does not collect personal data on its own.

FAQ

Does it read PDFs from Google Drive, Dropbox or S3? Use a direct public download link (for Drive, a uc?export=download link). Private links that need login cannot be fetched.

Can I send the PDF content instead of a URL? Not yet; upload the file somewhere with a direct link or an Apify key-value store public URL.

How fast is it? Typically a few documents per second for normal reports; large files depend on size and the number of pages.

Which output should I use for LLMs? markdown keeps headings, which helps chunking; pages helps with citations.

Feedback

Found this useful? A quick review on the Store helps others find it. Issues or feature requests? Open an issue and I'll reply fast.